Activation steering is a lightweight alternative to fine-tuning for aligning and personalizing large language models (LLMs). Prior work such as Bi-directional Preference Optimization (BiPO) learns dense steering vectors from preference data. However, dense interventions can entangle latent factors due to neuron multi-semanticity, hurting stability and fine-grained control, thus limiting their effectiveness especially in cultural alignment, where closely related values and behaviors (e.g., across Middle Eastern cultures) must be distinguished. We propose Yet another Policy Optimization (YaPO), a reference-free method that learns preference-optimized steering vectors in the sparse latent space of a Sparse Autoencoder (SAE). While the learned vector is not explicitly constrained to be sparse, it acts over sparse SAE features, providing a more disentangled and interpretable intervention basis. By optimizing sparse codes, YaPO yields more disentangled and interpretable steering directions. Across cultural alignment and several alignment-related behaviors (e.g., hallucination reduction and jailbreak resistance), YaPO converges faster and is more stable than dense steering baselines while improving performance. Importantly, YaPO preserves general knowledge, with no measurable degradation on MMLU. Overall, our results show that YaPO provides a general recipe for efficient, stable, and fine-grained alignment of LLMs, with broad applications to controllability and domain adaptation.

Jais 2 is a family of open-weight, Arabic-centric large language models designed for strong Arabic performance and broad cultural coverage. The family includes, to our knowledge, the largest open Arabic-centric LLM trained from scratch, with 70B parameters, alongside an efficient 8B model. A custom Arabic-centric vocabulary supports efficient training and inference, while an optimized architecture and training recipe improve compute efficiency. Jais 2 achieves the highest scores on the Open Arabic LLM Leaderboard v2 (OALL2) and AraGen v1 among the evaluated open models, while remaining competitive on English benchmarks. It also obtains strong results in Arabic poetry, dream interpretation, translation, and summarization. On the IslamicQA domain, Jais 2-70B achieves the second-highest overall average among the evaluated large open models, while attaining the highest scores on PalmX and the held-out in-house IslamicQA benchmark. We release both model sizes as open-weights on Hugging Face under a commercially permissive license. Jais 2-70B is also available as a chat application on the Web, iOS, and Android, reaching inference speeds of up to 2,000 tokens per second on Cerebras hardware.

We introduce Nile-Chat-4B, 3x4B-A6B, and 12B, a collection of LLMs for Egyptian dialect, uniquely designed to understand and generate texts written in both Arabic and Latin scripts. Specifically, with Nile-Chat-3x4B-A6B, we introduce a novel language adaptation approach by leveraging the Branch-Train-MiX strategy to merge script-specialized experts, into a single MoE model. Our Nile-Chat models significantly outperform leading multilingual and Arabic LLMs, such as LLaMa, Jais, and ALLaM, on our newly introduced Egyptian evaluation benchmarks, which span both understanding and generative tasks. Notably, our 12B model delivers a 14.4% performance gain over Qwen2.5-14B-Instruct on Latin-script benchmarks. All our resources are publicly available. We believe this work presents a comprehensive methodology for adapting LLMs to a single language with dual-script usage, addressing an often overlooked aspect in contemporary LLM development.